Two papers accepted to EMNLP 2025 on efficient model pruning and reinforcement learning with LLM-derived priors!